TY - CONF

T1 - Using multiple computer-predicted structures as MR models in protein crystallography: application to antiviral mini-protein LCB2.

AU - Корбан, Светлана Андреевна

AU - Михайловский, Олег Владимирович

AU - Лузик, Дмитрий Александрович

AU - Гуржий, Владислав Владимирович

AU - Лёвкина, Александра Денисовна

AU - Харьков, Борис Борисович

AU - Скрынников, Николай Русланович

PY - 2024

Y1 - 2024

N2 - LCB2 is an engineered mini-protein that binds to spike protein of SARS-CoV-2 and thus neutralizes the virus [Cao et al. Science 2020; 370:426]. We have solved the crystallographic structure of this mini-protein to a resolution of 2.1 Å. LCB2 has no significant sequence homology to any protein in the RCSB database. This prompted us to test a number of computer-predicted models in the role of molecular replacement (MR) model to solve the structure of LCB2. We found that AlphaFold2, Rosetta, MultiFold, RoseTTAFold and trRosetta (Yang Lab) all produce successful MR models leading to the desired structure. On the other hand, older algorithms such as Phyre2, Swiss-Model and I‑Tasser failed to produce a viable solution. The five successful MR models showed a significant amount of structural variability; at the same time the five respective final structures are in excellent agreement with each other (within 0.30 Å). This result characterizes the precision (and by extension the accuracy) of our medium-resolution structure. The convergence of the structure calculations starting from different MR models can be conveniently visualized using the principal component analyses. Generally, we advocate the use of multiple computer-generated MR models since this approach helps to elucidate different elements of the structure. For instance, in our study of LCB2 the side chain of residue R49 has been successfully resolved using some of the MR models, but not the others. This work was in part supported by the SPbU grant AAAA-A16-116102010033-6.

AB - LCB2 is an engineered mini-protein that binds to spike protein of SARS-CoV-2 and thus neutralizes the virus [Cao et al. Science 2020; 370:426]. We have solved the crystallographic structure of this mini-protein to a resolution of 2.1 Å. LCB2 has no significant sequence homology to any protein in the RCSB database. This prompted us to test a number of computer-predicted models in the role of molecular replacement (MR) model to solve the structure of LCB2. We found that AlphaFold2, Rosetta, MultiFold, RoseTTAFold and trRosetta (Yang Lab) all produce successful MR models leading to the desired structure. On the other hand, older algorithms such as Phyre2, Swiss-Model and I‑Tasser failed to produce a viable solution. The five successful MR models showed a significant amount of structural variability; at the same time the five respective final structures are in excellent agreement with each other (within 0.30 Å). This result characterizes the precision (and by extension the accuracy) of our medium-resolution structure. The convergence of the structure calculations starting from different MR models can be conveniently visualized using the principal component analyses. Generally, we advocate the use of multiple computer-generated MR models since this approach helps to elucidate different elements of the structure. For instance, in our study of LCB2 the side chain of residue R49 has been successfully resolved using some of the MR models, but not the others. This work was in part supported by the SPbU grant AAAA-A16-116102010033-6.

M3 - Abstract

SP - 11

T2 - Chinese Biophysics Congress 2024

Y2 - 25 July 2024 through 28 August 2024

ER -